Supporting Agile Software Development by Natural Language Processing - Trustworthy Eternal Systems via Evolving, Software Data and Knowledge - page 97

constituent parser (a parser that provides a hierarchical structure in which
smaller parts are combined into larger parts called phrases, e.g. a noun phrase
denoted NP) to the user story shown in Figure 2, we obtain the syntactic parse
tree shown in Figure 4.

S

PP NP VP

IN NP- ROLE PRP VBP S

As DT NN NN NN NN I want VP

a string manipulation library user TO VP

to AUX NP- GOAL

have DT JJ NN

a fancycase method

Fig. 4. A syntactic parse tree for the sentence “As a string manipulation library user, I
want to have a fancycase method” from the example user story (punctuation omitted,
abbreviated for space reasons). The tree is enriched with the target entity information
(in bold face).

The same process can be applied to the artifacts: parsing the commit mes-
sages, the code comments, etc. Based on the syntactic structure, a classifier
can be trained that determines the constituents that encode ROLE , GOAL or
BENEFIT of a user story (indicated in bold face in Figure 4) and similarly
of the artifacts. This leads to a possible structured instance representation that
can be exploited, as discussed in the next section.

3 Approach

In order to establish a connection between the user stories on one side and the
artifacts on the other site, we need a mechanism to associate them based on
their similarity. In this section, we outline our proposed approach to apply NLP
to artifacts obtained during agile software development in order to support the
product owner's decisions.
To this end, we propose a two-step approach as depicted in Figure 5: In
the first linking step, we establish connections between user stories and the
development artifacts (cf. Section 2.1). In the second information aggregation